Ramsri Goutham Golla joins SlatorPod to talk about his creator journey in AI SaaS (software-as-a-service), AI consulting, and AI courses.
Ramsri discusses his shift from software engineer in Silicon Valley to entrepreneur in India with AI SaaS apps Questgen.ai, Supermeme.ai, and Supertranslate.ai. He shares how a “build in public” strategy can help inspire people to follow your journey.
Ramsri talks about his blended approach to building products, marketing, and driving his social media channels. He touches on his most recent project, Supertranslate, which is powered by OpenAI’s Whisper for its one-click subtitle generation.
Ramsri shares how to build quickly on large language models like Whisper on a serverless GPU infrastructure and by partnering with developers who can build the front end. He gives his thoughts on the impact of ChatGPT on micro SaaS businesses and companies that have built on GPT-3.
Subscribe on Youtube, Apple Podcasts, Spotify, Google Podcasts, and elsewhere
Ramsri outlines some of the movers in the data science space that have propelled innovation, from serverless GPUs and open-source machine learning to vector search companies. He talks about how to form a startup in India, including tips on logistics, operations, hiring, and whether to raise funding or not.
The pod rounds off with Ramsri’s plans for the future, where he aims to eventually settle on one project full-time and sell off the others.
Transcript
Florian: How does one become one of the most interesting Twitter accounts out there?
Ramsri: Some people have it easy with their following, but I do it almost like a job, which is every day I wake up and allocate two hours to content since I have been on this creator journey. We can go deeper into that later, but short answer is a lot of hard work and I repurpose content on LinkedIn, Twitter, Medium, and YouTube.
Florian: Let us go a little bit further into what you do. You posted recently a very nice graph of all your different activities. You call yourself a creator and your work broadly falls into three practices: AI consulting, AI courses and AI SaaS. Under AI SaaS you started something called Supertranslate.ai, so give us a brief tour of your online presence and some of the projects that you are running.
Ramsri: I call myself a creator for the sake of no other better word that does not give you a salary every month. It has been almost two and a half years since I saw a monthly salary. It was by chance that I got into this because two and a half years back I had a job. I thought I could do better than that. I started this journey and I realized there were multiple things that I loved and I wanted to blend everything in, so in short, I do three things. One is I have three small AI micro SaaS, and two of them are in the $1,000 plus monthly recurring revenue range. One of them just got started, Supertranslate.ai. Besides that, I teach courses on Udemy as well as on my own website, learnnlp.academy. I mostly focus on natural language processing, GPT-3, all this stuff. The third thing that I do is AI consulting. Once in a while, whenever there are interesting projects, I take up consulting in the space and deliver them. In general, three things that I do apart from this content creation actively on LinkedIn, Twitter, Medium and YouTube.
Florian: What is your background, like university, degrees, prior jobs? You are commenting very expertly on AI and you are one of the most interesting accounts to follow. Do you have a background in this specific field or something adjacent which makes it a little easier for you to catch up on all the technical stuff that is going on?
Ramsri: I do have a background in this. Primarily I come from signal processing, image processing background. I did my master’s from Arizona State University in 2011 to 2013. Then I worked in the Valley pre-deep learning as a neural network era engineer. Then during 2015, 2016, two big revolutions happened. One is the deep learning revolution. Second is the mass online learning coursework, Coursera, Udemy, Udacity, all these things took off. I upskilled myself into all these areas, mostly in the AI space. I first learned about self driving car technology and then slowly moved on to other generative AI, as people call it today. Back in 2018, I pitched in Hong Kong and a few other places about generating AI art. It has gotten more popular this year and last year, but that is the short journey. I do come from this background, but at school the tech was not up there and I learned quite a lot online. In 2018 I took a massive turn from coding all this stuff to learning the business side and creator side. I made a dramatic shift from Silicon Valley to India. After eight years in the Valley, I moved back to India to try and do entrepreneurship and learn about other things like creator journey, consulting, and courses.
Florian: When you come with a technical background like yours and you are seeing all these evolutions and revolutions and disruptions happening, what level of overwhelmedness, for lack of a better term, do you experience? Or do you feel like this has all been a long time coming and it is the logical next step now?
Ramsri: I would say it is definitely overwhelming because I have been in and out almost four or five years in the deep learning AI space but even today it is hard for me to keep up with everything. There is speech, there is image, there is LLM text. Then you have Stable Diffusion, DALL-E in the image side. Then you have speech synthesis companies as well as Whisper models. It is definitely overwhelming for me to keep up with all these things. Sometimes I do focus on the products that I am working on or the startups that I am building. Mostly I focus on text and voice, not too much on the image space for the lack of keeping up with everything. It is extremely hard even for someone like me who has been in the space and building smaller systems with open source models. I have built these text generation apps previously, but even for me, it is extremely hard to keep up.
Florian: You are very transparent because what you do is called build in public. Tell us more about what is build in public and why would somebody want to do that?
Ramsri: I approach build in public this way. If I had a strong network in the localization industry and I build amazing technology, it is easy for me to connect with people and launch something. Although, if you see all my apps, one is catered for memes, one is for making quizzes in edtech and the other thing is for the localization industry. If I do not have a network of my own, the other way for me to hold a beacon on the internet is by saying that I am doing this and relevant people can find me either through SEO blogs or Twitter, LinkedIn, et cetera. I did not know the term build in public. I thought if I did not have the network, it was natural for me to post and attract the people who have the network to join me. On the customer side, people can discover that this app exists and on the other side, cofounders who have complementary skills can find me. In this journey, it became very natural for me to say that I am struggling on this part, then I put together this app and this is live now. That became a daily routine for me and in the recent years, build in public momentum has caught up. I feel blessed to have naturally found this because I do not have an offline network in the verticals that I am working in because coming from software engineering background, I just do coding and AI related stuff.
Slator Software-as-a-Service (SaaS) Localization Report
Florian: You are very transparent with the numbers. What build in public people do is post daily users, daily revenue even at very low levels. Does that put a lot of pressure on you on the one hand? And is that what is giving this amplifier effect? If people are transparent, then you gravitate towards following these accounts as opposed to accounts that just repost PR and marketing.
Ramsri: I will tell you two approaches that I do. The first one was growing up in India I always wanted to know where someone was working. Microsoft or Google? I wanted to know their net worth or what they make, what they take home. If I did not know them, there was no way for me to ask their salary upfront or how much they make. You do not know what anyone makes and you assume from their life. If you are better informed, you can make better choices, whether you want to make that kind of money with that work style. Secondly, you can inspire other people to follow your journey, whether there is money or not. Let us say I am digging and there is no gold here, I can at least tell other people that this is waste of time and to go find another place. Anyway, I am not VC-backed to hide or show growth metrics to others. I will take my small portion and let others who have energy and effort and hard work also take some. Why cannot people tell what they make? Why is it a social norm to not say it? I started from that perspective to, I made only $1,000 from this course in two years, but this course gave me $20,000 and this is what I made in SaaS. If I am not taking a FANG job, I wanted to tell people that there is another way you can make $1,000 from course, $1,000 from SaaS and $1000 from consulting and survive. I am very transparent that I do not even advocate for people to take my journey. One, it is mentally taxing, especially if you have family or other obligations that you need to fulfill, either paying loans, et cetera, so I try to give both sides of the coin. If someone is chasing freedom and they are afraid to take that first leap, I can say that I am doing this with a kid, toddler and maintaining a family living in India. The other thing is I want to show that people living in India can make $150 an hour or $200 an hour with consulting. A lot of times if you do not say it no one gets inspired and they sometimes take less deserving opportunities. If you are transparent, you will either inspire people or you will let people make better choices for themselves that maybe this is not the right thing for them if they want a fixed income and other things.
Florian: What social channels have you chosen to push? It cannot be everywhere. Where would you see the greatest return on time invested?
Ramsri: I first saw my greatest ROI on LinkedIn. Now I am at 46,300 followers and I launched my first course there and got good ROI. Recently, I do not see much growth on LinkedIn. Even Twitter people say the same thing once they have high growth. I am very aware of platform risk and growing audience on only one platform, so I diversified into Twitter and I am at 5950 followers. I am slowly growing on Twitter, YouTube, and Medium. I am trying to diversify into all these channels. I look at it this way. Daily content should go on two platforms max. In terms of energy for me, LinkedIn and Twitter is that. But you need some long term strategy, like SEO or discoverability on the internet if you want to simplify it. You need a blog or YouTube. I try to do both sometimes. The reason why I do this is not just to grow my SaaS or other platforms, it is if people find anything that I can offer value on, it is return on investment for me. This is sometimes without any call to action or deliberate outcome even. I love creating content I know will reward. My biggest strategy has been this, the world is evolving such that whatever job you hold should be internet-facing if you want to thrive for the next 10 to 20 years.
Florian: How do you balance building products and marketing? Being on social media, how do you balance that? Is that very deliberate? Is it more how you feel on each day?
Slator 2022 Language Industry Market Report
Ramsri: It is a blend. Some people follow a marketing week and building week structure, but generally so far I have not done anything like that. For me, everything is blended. For example, if I have implemented most features for SaaS and I have time, I go and update a course or create a new course. When I am bored of courses, I work on SaaS updates because people will be emailing and there will be some consistent themes that people are asking for. The other thing that I do is whatever I encounter, whether it is understanding payments with Stripe or how to do something in no-code tools like Bubble.io, whatever I struggled with today, I will post it tomorrow into a blog post or a Twitter thread or LinkedIn post. What I realized is that if with ten years of experience I struggle through something, then I know that at least beginners have something to learn from that. Although, I might look stupid sometimes not understanding the taxes in Stripe, I still go ahead and ask. I realized recently that even if you have a SaaS operating out of US and have nothing to do with Europe, you still have to collect and report taxes if you have one European customer even for some of the countries. You have to pay VAT, at least register online and pay and that was not obvious to me. There are so many things like that, even Whisper. Once I used it in production for some outputs and it was so bad that you cannot even imagine how these two, three sentences were randomly generated by Whisper. I tell people that we found five videos which were fine to listen to, but Whisper did very bad on them. I tell people about things that only a practitioner would know.
Florian: Let us talk about Supertranslate. What is Supertranslate? Who is using it? How do you see traction? Where do you see traction?
Ramsri: In short, Supertranslate, as of today is a one-click subtitling app. No matter whether your audio or video is in 100 plus languages, you can subtitle in English in one click. People do not understand the whole of it. Coming from localization industry, you would be aware that there are guidelines. It should not exceed certain characters and the split should happen not abruptly, but where there is a conjunction, so that it is smooth. The other thing is speaker diarization and identifying different speakers or word level timestamps. Subtitling itself has quite a lot of nuances beyond getting the transcription. How do you split the transcription perfectly with split length taken into account and other things? Once Whisper came out, we realized that there is a new wave of subtitling that you can do with Whisper. If you wanted to convert any video or audio from French to English, you previously had to do French speech-to-text, then a translation, and then you would get the subtitles. What Whisper did, which was revolutionary, was one-click translation directly from any language to English without any intermediary translation engine. It was about time to leverage this one-click direct translation as well as the accuracy of Whisper and build on top of this. The end goal is much bigger, which is to build one-click motion graphics for short form creators. For example, you might have seen these videos where text appears as motion graphics popping up, zooming in, sliding, things like that. Can you do that in one click? If you give me a short form video that you are posting on Instagram, I can give an output video with subtitles that have Whisper’s accuracy, but with animations that are overlaid. You do not need to edit for 2 hours to get those kind of animations, rather you have one-click motion graphic subtitles.
Florian: Whisper is an open source model, so how does it do the quote-unquote translation? Is it all baked into the giant model? Or do they have an additional machine translation step? Do you know how that is done?
Ramsri: When I say translation, Whisper is trained for any language to English subtitles, so it cannot do other combinations. It can take in any audio from 100 plus languages and get accurate English subtitles. That is about it. There is no explicit translation engine, the neural network itself is trained such that you have Hindi audio or French audio, then you get English transcription. It is trained in such a way that you cannot point out that this is a translation engine. It is almost like a human speaker who hears French and writes English.
Florian: You spun up this front end micro SaaS within a few weeks. How does one do that? What is your stack like for payments and things like that? How can you build something so quickly when something like Whisper comes out?
Ramsri: One of the biggest things is hosting Whisper to some scale. Whenever a new model comes up, especially with these language models, it is hard because you need a GPU to host. One of the biggest revolutionary things that happened in the last year, is serverless GPUs. Serverless GPU simply means you call and only pay for the compute that has being used. Let us say you used 1 minute of GPU time, you only pay a few cents. Now no matter whether Whisper or some other model came up, instead of you hosting that for $500 and paying $500 bill for one GPU, you can instead use serverless GPU where you only pay per usage. One of the biggest innovations was quickly taking any open source model and building an API on top of it with serverless GPUs to host it. Then you partner with developers who can build the front end very quickly. Since I have built SaaS at least three or four times, I know the fastest tech stack to pick from serverless GPUs to AWS to front end. I test it out and launch by building in public on LinkedIn and Twitter. That is where I get my beta users. Who are the people who are currently using it? Simply anyone who wants subtitles. It could be YouTubers. There are many non-native YouTubers who put out content in Hindi or Spanish, et cetera. They use Supertranslate and in one click they get the subtitles in English. They probably them edit a little bit and then get the SRT file. We launched in probably one month, one and a half months time after initial tests and now it is at $200 and 14 or 15 paying customers.
To be honest, we do not publicly post too much about Supertranslate. Simple reason is that our tech implementation is lagging behind because all our developers are part time. We could not move at the developer speed that we wanted to. We wanted to keep low and build more features like smart splitting and speaker diarization because Whisper does not support speaker identification natively. You have to merge various transcription services onto Whisper and do a blend of things, et cetera. There is some R&D work happening there. We are laying low and hope people do not discover us. It is not fully ready yet and that is why we are moving slowly. There is one thing we are doing currently, which is one of our cofounders is a localization expert sitting out of UK. He has been traditionally running this localization agency with regular translators, doing this stuff. We are very confident that if we have the right platform we can first integrate with him and build it there so that we have a critical audience and feedback loop established right there. Then we are seeking out feedback from the general public as well. We are building out a system in the localization industry. You might be aware that you need to assign tasks to certain people and you need to have multiple checks and so it is almost like you need to have half project management and half Whisper or whatever it is.
Florian: 100%. It is a project management industry.
Ramsri: Exactly, so we are building that out and we are in the initial phases and that is what we are moving towards. It will be sometime before we can make it open to general public at that scale.
Florian: Back to serverless GPU, how does that differ from AWS in the past? It was great for startups in the past 15 years, so why is there an opening for serverless SaaS when there is something like AWS?
Slator 2022 Language Industry M&A and Funding Report
Ramsri: First of all, whenever people talked about serverless and compute, mostly they were referring to CPU because most of the tasks that you wanted to do were not compute heavy. Only deep learning and all these high GPU processing tasks are compute heavy. What happened was spinning up a GPU server, doing a certain task and closing it down because the deployment and hosting and all those packaging for GPU is slightly more complex than CPU. You can use intel CPU or any other CPU and run a task and get the job done. When it comes to GPU, there are so many layers because you need to work with correct kernels that are used with Nvidia GPU or any other GPU that you use. There is quite a lot of infrastructure complication that is going on with GPU. Whenever you have a task, you need to spin up a GPU or keep it ready, do the task and shut down or keep it low, apart from the memory of RAM, et cetera. It is a whole new ball game scaling GPUs and maintaining them serverless. That need was not there until the last one, two years and anyone building huge models, they had their own GPU servers. The short answer is that AWS also has GPUs. You can use those GPUs but you need to write custom code to do it. It is a 100k or 150k developers task to maintain these GPUs, use them and shut them down. Instead these serverless GPU companies have provided a layer which is almost like a UI/UX layer and you do not need an explicit developer to handle this. Rather any software engineer can use it, quickly spin it up, do the task, and get it done.
Florian: It is also because there is this explosion of micro SaaS like yours that need access to that compute without having to build this giant complex infrastructure and so we are going to see a lot more of these serverless GPUs.
Ramsri: What you said is perfect because now micro SaaS can pay per compute and be profitable, monetizable. At least SMB level, you do not go for dedicated GPUs, you want to use compute level. That also spun the demand. All these Stable Diffusion, image generation things added fuel to the serverless GPU fire.
Florian: Let us talk about ChatGPT. How do you see this impacting what you are doing? How is it impacting the narrative around building these applications on top of these powerful models?
Ramsri: ChatGPT is revolutionary. I can tell from personal experience that no other technology or tool has penetrated to the level. It was never heard of and in this span, which is two, three months, local newspapers are writing about ChatGPT. Local YouTubers are talking about ChatGPT. I am super impressed with the marketing because they could be a revolutionary AI tool, but this kind of marketing and viral growth loop is something I have never heard of or seen in the 10 years of my journey. Everyone who has played with it would understand, from writing code to marketing copy, it has been a disruption all over. To be honest a lot of companies, micro SaaS or companies who have built on GPT-3, have seen decline in their month on month growth because of widespread ChatGPT growth. In one way it did affect some of our SaaS, Questgen a little bit. One of the positives is that what I am building is in an edtech vertical. I am building this quiz generator, so although people can go ahead and create their own prompts, et cetera, it is not like you can come up with the best prompt and maintain one for true or false quiz, one for MCQ. At least you can be somewhat defensible, but time will tell. To be honest, it is definitely scary for a lot of SaaS builders who have built not in a vertical, but a regular SaaS. Even in vertical, a lot of people that I talk to have been using ChatGPT directly.
Florian: I am a very SaaS curious person, so I keep registering for services here and there. With this explosion of micro SaaS, is there something you can do other than signing up? Understanding what is useful tech and what is a front end with a nice landing page, but not much behind it.
Slator Pro Guide: Finding Growth in Adjacent Services
Ramsri: When it comes to micro SaaS, it is built by a handful of people and does only one thing very well. It does not diversify into everything. Their goal is not to build a billion-dollar company, so they are not attacking a full market, but rather a niche within that market. For example, I am not trying to attack the education sector at all. I am focused on creating quizzes, classroom assessments, and that is about it. With microSaaS sometimes people build landing pages, collect emails, and do nothing. Sometimes it is fair, sometimes it is not. Sometimes they even take the payment and do not deliver it. Sometimes they refund and sometimes they do not. There are definitely cons with micro SaaS which is not a high level of accountability. To be honest, we are working on multiple projects. It is divided time so you cannot grow and deliver at the level of a VC-funded dedicated team, so there are a few pros and cons. Usually I sign up and if it is beyond an email which says we are launching soon, et cetera, I do not usually go ahead and do that because I have seen it from zero to one so I know exactly how those minds operate. If they are saying it is in the pipeline, it might never get there. Although, Product Hunt allows you to launch only when you have a working demo, not just showcase landing pages, et cetera. My go to is going to Product Hunt because they have at least one feature working well. Then when I try it out and I am impressed, I usually try to get subscriptions if it is something that saves me time. One of the other things is I do a lot of curation. All the voice AI startups that are coming up, I write threads on that. From a creator perspective, it is not about one problem that I have, but rather I want to know how far they are developed so that I can mention the top ten tools whether they are in that or not. From that perspective it is useful for me.
Florian: Let us talk about startup scene in India. I am seeing a lot coming out in the AI dubbing and transcription space. Is natural language processing generally a big deal in India? What are some of the other areas related in the broadest possible sense to language that people are building in?
Ramsri: Data science is a big area in India that people are upskilling or have been involved in. I would say research happens in Silicon Valley and all around the world and some things are propagated. Only in the last three, four years, people have started collecting data and building their own models on open source tech stack/ Companies like Hugging Face, et cetera, also democratized that AA building now with serverless GPUs and other things. There are some few movers in this space that propelled the innovation as well. One is serverless GPU. One is open source ML like Hugging Face. Another is vector search companies providing semantic search, vector search, and vector embeddings, also as serverless or pay-per-use sometimes. With all these things, it has become easier to iterate and do R&D as well with Google Collab, et cetera. Google Collab is free where you can sign up and train your own GPU model on there for a couple of hours of compute for free. Still a lot of research happens outside of India and then it gets propagated here but at least in the last three, four years I have seen good number of companies being founded. I think you are familiar with Dubverse and other voice startups and large language model startups that are coming up. Let us see how far things go.
Florian: How do you form a startup in India? How do you incorporate? How do you hire? How do you scale? I want to know the nuts and bolts.
Ramsri: It is not too bad. Probably a two-week to one month process. If you are building something equivalent to C-corp in India, you need to have at least two founders and bank account, et cetera. You can almost form a company online. Forming a company is not the toughest part. There are several other things which is you need to do quarterly filings if you cross a certain threshold, et cetera. The threshold is not too big when you operate even at medium scale. The logistics, et cetera, are definitely there and it is the same everywhere. To be honest, I do not think the operations part is complex. If you have the money you can hire a chartered accountant and there are companies which can do the task for you. A few things that I have noticed is when it comes to raising VC funds and what salaries you can draw as a founder, a lot of things need to improve. To be honest, one reason why I have not raised funding is I make more than what an average seed or series A founder makes. For me it never made sense and I will tell you one more catch, which is I have registered as a Delaware C corporation. Anyone can register as a US Delaware C-corp without being and staying in the US. It is expensive to operate, I will put that upfront, but I am the type of guy who can pay ten X for an easier process and to do everything online. In the US I could register on my own, sending details from here. I never had to sign anything in ink. Everything was online.
I would say hiring is easier in the sense that at least in the initial phases, there are interns who will work for $60, $70 monthly, because you have 1.3 billion people. Good startups do not exploit it and the quality is what you get even. Let us say you are doing data labeling, you can hire an intern for $100 and let them do data labeling for a month even. When you get to the competitor level, hiring is tough in the sense that I would say sometimes it is easier to find good people in Silicon Valley than in India. People who are good here will ask for 150k, 300k sometime. Some people have hired from Silicon Valley at 100k. Staying remote then paying engineers here. It is asymmetric in the sense that top engineers, what they make, the lifestyle they have is definitely asymmetric. If you make close to 200k in India, you almost live like a 800k engineer in Silicon Valley because you can afford someone to cook food, take care of household, and you can afford luxuries. If you are in the top brass, you enjoy asymmetric benefits even when compared to counterparts in US. That is what I have seen. With 1.3 billion people the competition is also cut throat. That is one of the things that a lot of people say. That is why everyone pushes you to engineering or something like that, not because they are against other arts. With 1.3 billion people and resources like this, you have to cut corners. The best you can do is equip yourself, at least on a personal level, upskill yourself and be better.
Florian: Is it quite different between the cities? Now you are in Hyderabad, is it very different than Mumbai or Delhi or has it flattened out?
Ramsri: It is very different in the sense that Bangalore is like Silicon Valley where things are happening. Every other guy that you talk to might be into Web3, AI or full stack development, et cetera. There are always jokes about if you have your laptop open with ChatGPT and a Web3 t-shirt, then a VC might approach you if you are sipping your $20 coffee. Fortunately in Hyderabad, at least where I live, the startup scene exists, but it is not huge. I love it in the sense that I am bootstrapped. I am isolated from everybody. I do not have a big ecosystem where have mentors or anybody. I build in silos and build in public online. I am not too concerned and it is harder to raise venture capital or hire folks in Hyderabad than in Bangalore. If I had all these troubles, which is raising venture capital as well as hiring team, I would definitely feel left out staying in Hyderabad as opposed to Bangalore. Bangalore is like Silicon Valley and Hyderabad you can think of like Newark or sometimes Texas where there is some scene, but not too much.
Slator Machine Translation Expert-in-the-Loop Report
Florian: We have discussed the multiple projects that you are running, so what about 2023 and beyond? What is the plan? Are you going to settle on a single project full time or do you enjoy having all these different balls up in the air?
Ramsri: The ideal thing is to settle on one project and do it full time. A few things need to happen before that. I need to sell the other products which I do not think at the moment I am ready to. I enjoy working on them. Most of the people who want to buy it at five times ARR or something like that, it does not make too much sense for me. The reason why I want to work on one product in the long term is sometimes context switching hurts quite a lot. I need to maintain three GitHub repos for these three products. If a client pops up, I have to work on their stack, their AWS. For courses I need to work on creating and recording them. To be honest, it is taxing. The other thing is, if I raise VC money and go forward, you have to be committed for at least four to seven years to work on the project and be there. We want to go down that route, probably when we are making $10,000 MRR. If there is some product validation market fit that happens, then it is worth dedicating our next four to five years on that. Now micro SaaS is such that if tomorrow I am not happy with it or I do not have the time, I can go list it and sell it at standard pricing if I want and exit out of it. For VC, it is not easy to get an exit because you have to align multiple things, which is your company has to be on high growth curve as well as be a good acquisition target for somebody and it should give good returns to VCs. Let us say Questgen is making $1,500, I can sell it at 100k and walk away with that 100k and I am the only sole decision maker. I do not need to do anything and I feel comfortable that way. Now I love working on all of them because no one pushed me to even do them in the first place and all of them are profitable so I am not losing money on any of them. Through the micro SaaS journey I have built such that even at $100 MRR it is profitable because I use only pay-per-use serverless GPUs, pay-per-use vector search and all these things.
Florian: That is incredible. We are at the start of a revolution. If you can spin up something that is workable for $100 or $200 per month that somebody can use, wow.
Ramsri: I am not even a full stack developer. I come from AA background but I learnt Bubble no-code technology and built a UI, UX and launched Questgen all on my own. I did not have to spend tens of thousands of dollars on developers who do back and front end. I operated until recently on a $29 stack with Bubble. I was maxing out recently the compute so I upgraded to $129 plan. I was paying for GPT-3 cost. If one month no one uses it, I get zero dollar bill but if someone uses it, maybe I get $200. I was paying only $29 per month to Bubble and everything was integrated, Stripe, et cetera, so anyone could come in and pay. I always say, this is a lifestyle and whether this lifestyle is ideal for you at the stage that you are in. One needs to evaluate whether it is right for them. I moved from Silicon Valley in 2018. Not even for a split second I thought, I should have stayed there. I m totally fine here. I only think that, if I do two X MRR on this SaaS, I would be more free in terms that I do nOt need to take consulting.
Florian: Where do people find you on Twitter and LinkedIn? Ramsri: You can look me up. Ramsri Goutham Golla and my profile picture, name, description is almost the same on all the apps. It is very hard for you to miss if you Google me.